# Zero-shot Transfer

OpenVision is a fully open-source, cost-effective family of advanced vision encoders for multimodal learning.

ViCA2 is a 7B-parameter multimodal vision-language model focused on video understanding and visual-spatial cognition tasks.

BLIP is a unified vision-language pretraining framework, excelling in vision-language tasks such as image caption generation

SigLIP 2 ViT model, containing only the image encoder part for image feature extraction, trained on the WebLI dataset.

A Vision Transformer model based on SigLIP 2 architecture, designed for image feature extraction and pretrained on the webli dataset.

A vision Transformer model based on the SigLIP 2 architecture, designed for image feature extraction, pretrained on the webli dataset

Vision Transformer model based on SigLIP 2 architecture, designed for image feature extraction, trained on the webli dataset

ViT image encoder based on SigLIP 2, designed for timm, suitable for vision-language tasks

A ViT image encoder based on SigLIP 2, employing global average pooling with the attention pooling head removed, suitable for image feature extraction.

Vision Transformer model based on SigLIP 2, designed for image feature extraction, pre-trained on the webli dataset

ViT model based on SigLIP 2, focused on image feature extraction, trained on the webli dataset

A vision-language model pre-trained on the COCO dataset, excelling in generating accurate image descriptions

ConvNeXt-Large image encoder based on CLIP architecture, fine-tuned on the LAION-2B dataset, supporting 320x320 resolution image feature extraction

ConvNeXt-Large image encoder based on the CLIP framework, trained on the LAION-2B dataset, supports visual feature extraction

CogACT is a novel advanced Vision-Language-Action (VLA) architecture derived from Vision-Language Models (VLM), specifically designed for robot manipulation.

CogACT is a novel advanced Vision-Language-Action (VLA) architecture derived from Vision-Language Models (VLM), specifically designed for robot manipulation.

CogACT is a novel Vision-Language-Action (VLA) architecture that combines vision-language models with specialized action modules for robotic manipulation tasks.

AIMv2-Large-Patch14-Native is an adapted image classification model, modified from the original AIMv2 model to be compatible with Hugging Face Transformers' AutoModelForImageClassification class.

SAM 2 is a foundational model for promptable visual segmentation in images and videos developed by FAIR, supporting efficient segmentation through prompts.

SAM 2 is a foundational model for promptable visual segmentation in images and videos developed by FAIR, supporting universal segmentation tasks through prompts.

SAM 2 is a foundational model for promptable visual segmentation in images and videos developed by FAIR, supporting efficient segmentation through prompts.

This model is a fine-tuned version of microsoft/Florence-2-large, trained on a subset of 40,000 images from the Ejafa/ye-pop dataset, with annotations generated by THUDM/cogvlm2-llama3-chat-19B, focusing on image-to-text tasks.

AM-RADIO is a visual foundation model developed by NVIDIA Research, featuring an aggregated architecture for unified multi-domain representation, suitable for various computer vision tasks.

RADIO is a vision foundation model developed by NVIDIA Research, capable of unifying visual information across different domains for various vision tasks.

This is a fine-tuned version of the microsoft/Florence-2-large model, trained on a subset of 38,000 images from the Ejafa/ye-pop dataset, using CogVLM2-generated annotations, focusing on image-to-text tasks.

This model is released under the Apache-2.0 license, with specific details currently unknown.

FashionCLIP is a vision-language model based on CLIP, specifically fine-tuned for the fashion domain, capable of generating universal fashion product representations.

Chronos-T5 is a pre-trained time series forecasting model based on a language model architecture. It converts time series into token sequences through quantization and scaling for training, making it suitable for various time series forecasting tasks.

ZoeDepth is a vision model for monocular depth estimation, fine-tuned on the KITTI dataset, capable of achieving zero-shot transfer for metric depth estimation.

A multilingual web register classifier based on fine-tuned XLM-RoBERTa-large, supporting text classification tasks in 100 languages.

This is an open-source model based on the Apache-2.0 license; specific functionalities depend on the actual model*

Open-source model under Apache-2.0 license (specific details unavailable)

BLIP is a unified vision-language pretraining framework, excelling in tasks like image caption generation, supporting both conditional and unconditional text generation

Open-source model based on Apache-2..0 license, specific functionalities depend on the actual model type

No model information available

BLIP is an advanced vision-language pretraining model, excelling in image captioning tasks by effectively utilizing web data through guided annotation strategies

BLIP is an advanced vision-language pretrained model, excelling in image captioning tasks, capable of generating accurate natural language descriptions based on image content.

No specific description provided for this model; unable to generate an introduction

BLIP is a unified vision-language pretraining framework, excelling in image caption generation and understanding tasks, efficiently utilizing web data through guided annotation strategies

LLaVA-Lightning-7B is a multimodal model based on LLaMA-7B, achieving efficient vision-language task processing through delta parameter tuning.